International Joint Ventures, Moral Hazards,and Technology Spillovers

Kar-yiu WongUniversity of Washington

Wing-fai LeungCity University of Hong Kong

January 11, 2001

Abstract

In this paper, we provide a theory to explain why a multinational corporationand a local ¯rm choose to form an international joint venture (IJV) and whythey decide to dissolve it sometime later. Our theory has two importantfeatures: moral hazard and technology spillovers. Because of moral hazard,a ¯rm may not provide the right e®ort to make an IJV bene¯cial to both¯rms. When an IJV is formed, technology spillovers allow both ¯rms to learnfrom each other. Under certain conditions, at least one of the ¯rms learnssu±ciently from the other ¯rm and decides to break away from the jointventure.

Thanks are due to an anonymous reviewer and the particiapants at the work-shop on \International Economics and Asia" held on July 19{21, 2000 at CityUniversity of Hong Kong. This project is partially supported by the Com-petitive Earmarked Grant for Research (No. 9040320).

c° Kar-yiu Wong and Wing-fai Leung

1 Introduction

There are many reasons why two ¯rms (or more) choose to form a joint ven-ture to produce a commodity. To increase the monopoly power is one reasonoften suggested. Another interesting reason is that a joint venture allowsthe partners to combine some of their advantages that are complementaryto each other, including ¯nancial capital and technological knowledge.1 Ofall the forms of joint ventures, those formed by ¯rms in di®erent countries,which are commonly called international joint ventures (IJVs), receive specialattention.For an IJV, the di®erence between a multinational corporation (MNC)

and a local ¯rm is often emphasized. According to the ownership-location-internalization (OLI) literature, a MNC that chooses to supply to a foreignmarket not through export but through direct investment must possess someownership advantages in order to overcome the obvious disadvantages of itsunfamiliarity local cultures and systems.2 We follow the approach of thisliterature, but focus instead on an option of a MNC that has not been coveredin the OLI literature: the formation of an IJV with a local ¯rm. Manystudies have pointed out that in forming an IJV, a MNC contributes someof its ownership advantages while a local ¯rm will make use of its knowledgeand familiarity of the local market and domestic cultures and systems.3

One feature of IJVs that has captured the attention of many people is thatthe life of a typical IJV seems to be short, in fact much shorter than the lifeof a typical and comparable wholly-owned subsidiary of a MNC (Beamish,1985; Franko, 1971; Kogut, 1988a and 1988b; Geringer and Hebert, 1991,and Li and Guisinger, 1991). In particular, Gomes-Casseres (1987) ¯nds thatalthough the liquidation of joint ventures is not very di®erent from wholly-owned subsidiaries (WOSs), the sellout rate of joint ventures (to one of thepartners or outsiders) is higher than WOSs. The paper is based on a sampleof over 5000 subsidiaries (either joint ventured or wholly-owned) of U.S.MNCs established between 1900 and 1975.4 The paper shows the share of

1See, for example, Beamish (1985), Benito and Gripsrud (1992), Blodgett (1991),Gomes-Casseres (1989), Hennart (1988), Kogut, (1988b), Lee and Beamish, (1995), Perl-mutter and Heenan (1986), and Reynolds (1984).

2See Wong (1995, Chapter 13) for a recent survey of the literature.3See, for example, Blodgett (1991), Kogut (1988b), Kogut and Singh (1988), and Leung

(1998).4Harvard Multinational Enterprise Project.

1

subsidiaries (of either form) that may be sold or liquidated by 1975, and anychange in ownership structure (from joint ventured to wholly-owned or viceversa) between date of entry and the date of liquidation or sale, or betweendate of entry and 1975. On the whole, the termination rate of IJVs is about30% while that of WOSs is only about 15%.5 Chowdhury (1992) furtherbreaks down the termination rates between joint ventures and WOSs intoseveral periods instead of the lump-sum data in Gomes-Casseres (1987). His¯ndings suggest that the average life of both IJVs and WOSs had consistentlydeclined from 1951 to 1975. Leung (1997) applies the proportional hazardmodel to study the duration of IJVs and WOSs by using a di®erent data setof U.S. MNCs. The results suggest that the IJVs have signi¯cant shorter lifespan than WOSs.Why does the life of a typical IJV seem to be short? If a MNC and a

local ¯rm agree to form a joint venture in the ¯rst place, why would theywant to dissolve it in the future? That the average life of IJVs is shorterthan that of WOSs suggests that the structure of IJV may be part of thereason. In fact, since the two ¯rms choose to dissolve a joint venture thatthey formed sometime ago implies that something must have happened afterthe formation of the joint venture.The objective of this paper is to explain why two ¯rms choose to form

an international joint venture and decide to dissolve it later. There are sev-eral features of our model: moral hazard, cooperation costs, and technologyspillovers. The presence of moral hazard means that one of the ¯rms maychoose to provide not su±cient e®ort to make an IJV attractive to the other¯rm.6 The cooperation cost, as suggested by Leung (1998), is another factor

5Termination is called \instability" in Franko (1971) and Gomes-Casseres (1987). Fol-lowing Franko (1971), there are three reasons why an international joint venture wouldterminate: (a) The joint venture is totally liquidated, i.e., the operation is terminatedor all the assets are sold or scrapped; (b) The ownership changes even though the jointventure is still operating, e.g., the joint venture is sold to the local partner or a thirdparty; (c) The foreign ¯rm (i.e., the MNC) may buy out the joint venture from the localpartner and create a wholly-owned subsidary. Gomes-Casseres (1987) points out that awholly-owned subsidiary can terminate if it is liquidated or sold to local ¯rms, or if theMNC sell part of the equity of the wholly-owned subsidary and change the venture formto a joint venture. The termination rate of international joint ventures is the numberof joint ventures terminated as a percentage of the total number of joint ventures. Thetermination rate of wholly-owned subsidaries is de¯ned in a similar way.

6For example, it was reported that Krohne, a German manufacturer of electromagnetic°ow meters, complained that its China partner in a joint venture was not working hard

2

that may encourage ¯rms to break away an existing IJV. If an IJV is formed,both ¯rms may learn from each other through technology spillovers, and withsuch technology improvement one or both of the ¯rms may choose to breakaway from the joint venture in the future. (See Benito and Gripsrud, 1992;and Parkhe, 1991, among others, for technology spillovers.)7 An implicationof the present theory is that an IJV can be formed and dissolved later evenin the absence of uncertainty or imperfect information.Sections 2 to 4 of this paper develop the model. For simplicity, we begin

with a single-period model. Section 2 focuses on a single ¯rm, whereas Sec-tion 3 turns to the case of duopoly, with a MNC and a local ¯rm competingin a Cournot fashion. Section 4 examines the formation of a IJV. How thelocal ¯rm chooses its e®ort in di®erent cases is explained. Section 5 explainshow the two ¯rms decide between forming a joint venture and producingseparately. Section 6 looks at a two-period model, which is used to explaintechnology spillover between the two ¯rms when an IJV has been formed,and to analyze the conditions under which the ¯rms would choose to breakup the joint venture later. The last section concludes.

2 The Model: Single Firm and Single Period

We ¯rst introduce a one-period model, which will later be extended to twoperiods to analyze intertemporal choice. Consider a homogeneous productand its market in an economy called \home." The demand for the productin the market is described by p = p(q); where p is the market price and q thedemand. It is assumed that p0(q) < 0 and that p00 is either negative or nottoo positive, where a prime denotes a derivative.8 The market is big enoughto support the two ¯rms to be introduced below.There is a single home ¯rm in the economy. Before the entry of the

foreign ¯rm, the home ¯rm is a monopolist. Its cost of production includestwo components: a variable cost that depends on the level of production,

enough. (\Enter in China: An Unconventional Approach," Harvard Business Review,March-April, 1997: 130{140.)

7In fact, technology spillover is often regarded as one major reasons for many developingcountries to attract foreign investment and joint venture between local and foreign ¯rms.See, for example, Miller et al. (1997).

8The latter assumption implies that the demand curve is concave or not too convex tothe origin so that the marginal revenue curve is downward sloping.

3

cx; where c is a positive constant and x is the level of production, and a\¯xed" cost which is independent of the level of output, f . This meansthat c is a constant marginal cost. The \¯xed" cost of production, whichis ¯xed in the production process, can indeed be changed by the ¯rm byspending di®erent levels of e®ort: Examples of \¯xed" cost are the costs ofestablishing good relations with the suppliers of inputs (factor owners andsuppliers of intermediate inputs, for example) and purchasers of its output(such as wholesalers and retailers), the costs of marketing and advertisement,and the costs of getting familiar with the local political, social, economic, andlegal systems. If we denote the e®ort made by the ¯rm by e; we can writethe \¯xed" cost of production as f = f(e); which satis¯es the followingproperties:

f(e) > 0; f 0(e) < 0; f 00(e) > 0: (1)

Based on condition (1), the value of ¡f 0(e) can be illustrated by scheduleFF in Figure 1. Note that ¡f 0(e) is positive, but declining.Since making e®orts is costly, we denote the cost of providing a level of

e®ort e by g = g(e); which has the following properties:

g(e) > 0; g0(e) > 0; g00(e) > 0: (2)

The ¯rst derivative of the function, g0(e); can be illustrated by a schedulelike GG in Figure 1. Condition (2) implies that the schedule is rising.For the sake of analysis, we assume that the home ¯rm takes two stages

to maximize its pro¯t. First, it chooses the optimal e®ort, e; and pays acost of g(e). The e®ort then gives a certain \¯xed" cost of production, f(e).Second, it chooses the optimal output level, x:We ¯rst describe the second stage. In this stage, both the e®ort e and

the \¯xed" cost f(e) are taken as given. The ¯rm chooses the output tomaximize its nominal pro¯t given by

µ(x; e) = p(x)x¡ cx¡ f(e); (3)

where the market equilibrium q = x has been used. Note that the nominalpro¯t does not include the cost of e®ort, which is paid for in stage 1. The¯rst-order condition is9

p+ p0x = c: (4)

9Note that an interior solution is assumed because the case with no output is notinteresting in the present paper. The second-order condition is assumed to hold.

4

Note that the ¯rst-order condition and thus the optimal output do not dependon the e®ort. That is because the e®ort is assumed to a®ect the \¯xed" cost ofproduction only.10 Let us use a superscript \m" to represent the equilibriumvalue of a variable when the home ¯rm is a monopolist; for example, xm isthe optimal output chosen by the ¯rm and pm is the resulting market price.The corresponding nominal pro¯t of the ¯rm is equal to

µ(xm; e) = p(xm)xm ¡ cxm ¡ f(e): (5)

In stage 1, the ¯rm chooses the level of e®ort to maximize its net pro¯t,which is the nominal pro¯t minus the cost of e®ort:

¼ (e) = µ (xm; e)¡ g (e) :The ¯rst-order condition is

¡f 0(e) = g0(e); (6)

where condition (5) has been used. Note that the output does not appear incondition (6), meaning that the optimal level of e®ort is independent of theoutput. Denote the optimal e®ort by em: The second-order condition is

¡f 00(em)¡ g00(em) < 0;which is satis¯ed because of conditions (1) and (2).Condition (6) and the optimal e®ort can be illustrated graphically. As

explained earlier, schedules FF and GG in the diagram represent ¡f 0(e) andg0(e); respectively. This means that the optimal e®ort is obtained from theintersecting point E between schedules FF and GG.

3 Rivalry between Two Firms

We now introduce a foreign ¯rm to the above model. Suppose that thereexists a foreign ¯rm, which is capable of producing the same product in thelocal economy and suppling it to the local market. For simplicity, we assumethat the foreign ¯rm has no production in its own country.11 For the time

10The result will be di®erent if the e®ort a®ects the marginal cost. For simplicity, thiscase is not considered in the present paper.11It may be due to, for example, prohibitive transport costs or a prohibitive tari®.

5

being, we consider the case of duopoly, with the foreign ¯rm and the local¯rm produce and compete in a Cournot fashion. Variables of the foreign¯rm are distinguished by an asterisk. In particular, it is facing a constantmarginal cost c¤ and a \¯xed" cost of production f ¤(e¤) after spending ane®ort of e¤ and paying a cost of g¤(e¤):The two ¯rms, however, have asymmetric positions. The foreign ¯rm,

being a newcomer in the local market, is in a disadvantageous position ascompared with the home ¯rm: it lacks good knowledge of the local polit-ical, economic, social, and legal systems, and, as compared with what ahome ¯rm has, usually has a less developed relationship with local suppliersand possibly wholesalers and retailers, and could have a cultural gap in theproduction process. Therefore a foreign ¯rm producing in another countrymust have some ownership advantages such as a more advanced technologyover the home ¯rms. To capture these two features, we make the followingassumptions:

(I) f¤(e¤d) > f(ed); where e¤d and ed (to be determined later) are the e®ortschosen by the foreign and local ¯rms when competing in a Cournot wayrespectively; and

(II) c¤ < c:

The ¯rst assumption is to capture the fact that the foreign ¯rm is in adisadvantageous position in producing the good in the local economy, whilethe second assumption represents the superior technology the foreign ¯rmhas with respect to the home ¯rm.To analyze the present case of rivalry between the ¯rms, we ¯rst have to

¯nd out how the entry of the foreign ¯rm may a®ect the home ¯rm's \¯xed"cost. The answer is nothing, which means that the e®ort chosen by the home¯rm is not a®ected by the presence of the foreign ¯rm. The reason, as shownby condition (6), is that the home ¯rm's choice of e®ort is not a®ected by itsoutput.As in the previous section, the competition between both ¯rms can be

analyzed in two separate stages. In the ¯rst stage, they choose the levelof e®orts, thus ¯xing the \¯xed" cost of production. They then competein output. Let us analyze the second stage ¯rst. With the levels of e®ortchosen, their nominal pro¯t can be written as follows:

µ(x; x¤; e) = p(x+ x¤)x¡ cx¡ f(e) (7)

6

µ¤(x; x¤; e¤) = p(x+ x¤)x¤ ¡ c¤x¤ ¡ f¤(e¤): (8)

Assuming Cournot competition, the ¯rst-order conditions of the ¯rms are:12

p+ p0x = c (9)

p+ p0x¤ = c¤: (10)

These are of course the reaction functions of the ¯rms, which are illustratedby schedules HH and FF, respectively, in Figure 2. The intersecting pointof the two schedules, N, is the Nash equilibrium. Let us denote the Nashequilibrium outputs of the home and foreign ¯rms by xd and x¤d; respectively.Based on assumption (II) and the assumption of a falling marginal revenuecurve, the home ¯rm, with a higher marginal cost, has a smaller marketshare. For a \stable" equilibrium, it is assumed that schedule HH is steeperthan schedule FF in a region close to point N.13 For a reason to be madeclear later, let us treat the home ¯rm's marginal cost c as a parameter anddetermine the output e®ects of a change in c. Di®erentiate (9) and (10) andrearrange the terms to give:24 2p0 + p00x p0 + p00x

p0 + p00x¤ 2p0 + p00x¤

3524 dxdx¤

35 =24 dc0

35 : (11)

Denote the determinant of the matrix by D = (2p0+ p00x)(2p0+ p00x¤)¡ (p0+p00x)(p0 + p00x¤) > 0; where the sign comes from the \stability" condition.Solving (11), we have

dx

dc=

2p0 + p00x¤

D< 0 (12)

dx¤

dc= ¡p

0 + p00x¤

D> 0; (13)

where the sign comes from the assumption that p00 is either negative or suf-¯ciently small in magnitude if positive. Conditions (12) and (13) imply thata drop in c will encourage the home ¯rm's production but discourage thatof the foreign ¯rm. This result is also shown in Figure 2. From (11), a dropin c shifts the home ¯rm's reaction curve to the right but the foreign ¯rm's

12This paper ignores the uninteresting case in which one or both ¯rms are not producing.13See Wong (1995). The \stability" condition is satis¯ed if the demand curve is not too

curved toward the origin.

7

reaction curve remains unchanged. The new Nash equilibrium point, N0, rep-resents a higher output of the home ¯rm but a lower output of the foreign¯rm.Conditions (12) and (13) can be combined to give the e®ect on the total

output:d(x+ x¤)

dc=p0

D< 0; (14)

which means that a drop in c will encourage a larger aggregate output of the¯rms.In the ¯rst stage, both ¯rms choose the optimal levels of e®orts to maxi-

mize their net pro¯ts

¼(e) = µ(xd; e)¡ g(e)¼¤(e¤) = µ¤(x¤d; e¤)¡ g¤(e¤):

Based on (6), the e®ort chosen by each ¯rm is independent of the output.Therefore, the ¯rst-order conditions are similar to (6). As we did earlier, weuse the superscript \d" to represent the equilibrum value of a variable in thepresent case; for example, ed and e¤d are the Nash equilibrim outputs chosenby the ¯rms, with the corresponding costs of e®orts given by gd ´ g(ed) andg¤d ´ g¤(e¤d); and the pro¯ts of the ¯rms given by µd ´ µ(xd; ed), µ¤d ´µ¤(x¤d; e¤d); ¼d ´ ¼(ed); and ¼¤d ´ ¼¤(e¤d); respectively. Note that becausethe chosen e®ort is independent of the output level, the home ¯rm choosesthe same e®ort in the present duopoly case as in the previous monopoly case,i.e., em = ed:As explained earlier, the disadvantageous position of the foreign ¯rm is

represented by a higher \¯xed" cost of production, i,e., f ¤(e¤d) > f(ed):

4 A Joint Venture of the Firms

In this section, we consider an alternative option for the foreign ¯rm toproduce the good in the local economy: forming a joint venture with thehome ¯rm. It is formed under the following conditions:

1. In the presence of a joint venture, both ¯rms would not produce sepa-rately.

8

2. The joint venture will have a marginal cost of c¤ and a \¯xed" cost ofproduction f(e) + Á; where Á > 0 and e is to be chosen by the home¯rm.

3. The joint venture chooses an output, X; to maximize the followingnominal pro¯t:

£(X; e) = p(X)X ¡ c¤X ¡ f(e)¡ Á: (15)

4. The home ¯rm receives a fraction s, 0 · s · 1, of the resulting pro¯t,and the foreign ¯rm receives the rest. The determination of s will beexplained later.

Let us explain these assumptions. Assumption 1 is a normal phenomenon.One reason is that both ¯rms are willing to stop producing separately as agoodwill gesture of cooperation. Assumption 2 represents the basis for thetwo ¯rms to cooperate in the present paper: the joint venture uses the foreign¯rm's advanced technology, in terms of a lower marginal cost, the home ¯rm'sbetter knowledge of the domestic market, leading to a lower \¯xed" costfrom e®ort. The variable Á refers to the cost of cooperation, which includesthe costs of coordination, communication, and possibly friction between the¯rms. For simplicity, we assume that Á is given exogenously. Assumption3 seems to be a natural one, but it has two implications. First, the levelof e®ort has yet to be determined. This is left to the home ¯rm to choose.Second, the cost of e®ort, g(e); has not been included in the nominal pro¯tof the joint venture. These two implications require further explanation.Since e®ort is a variable di±cult to quantify, measure, and verify, the

present model involves moral hazard by the home ¯rm.14 Even if both ¯rmssign an agreement that requires the home ¯rm to spend so much time onlowering the ¯xed cost, it is di±cult to determine or verify in court howmuch e®ort the home ¯rm has actually spent and what the cost of the e®ortis. In the present model with perfect information and no uncertainty, theforeign ¯rm can determine how much e®ort the home ¯rm will choose to

14Note that we assume that only the local ¯rm is responsible for the \¯xed" cost, andthus moral hazard occurs in that ¯rm only. Double moral hazard exists when both ¯rmssupply e®ort, but it is not needed for the results in the present paper.

9

spend under the present conditions.15 It is just that it is di±cult to requirethe home ¯rm to choose a particular level of e®ort or to verify that the home¯rm has chosen an e®ort di®erent from the one stated in an agreement.16

The foreign ¯rm knows well functions f(e) and g(e); but it is di±cult toverify in court what e®ort the home ¯rm has spent.Based on the above assumptions and description, we consider the follow-

ing three-stage game when a joint venture is formed. In the ¯rst stage, thehome ¯rm chooses the optimal level of e®ort. This in turn determines the\¯xed" cost of production, f(e): In the second stage, both ¯rms decide howto share the nominal pro¯t of the joint venture, and in the last stage, theychoose an output to maximize the nominal pro¯t of the joint venture. Ineach stage, each ¯rm is aware of what they will choose in later stages.We ¯rst examine the third stage. The joint venture chooses x to maximize

the pro¯t de¯ned by (15). The ¯rst-order condition is

p+ p0X = c¤: (16)

Note that this condition is similar to the ¯rst-order condition (4) for the home¯rm when it is a monopolist because in both cases there is only one singlesupplier. Let the optimal output be denoted by Xv; where the superscript\v" represents a variable in the presence of this joint venture. Furthermore,in both cases, the optimal output is independent of the level of e®ort chosenby the home ¯rm. There is a di®erence between the two cases, however,because the marginal cost of the joint venture is lower than that of the home¯rm, as stated in assumption (II). Thus (16) yields a higher output and ahigher pro¯t than (4) does. Denote the resulting nominal pro¯t of the jointventure by £̂(e):

£̂(e) ´ £(Xv; e) = p(Xv)Xv ¡ c¤Xv ¡ f(e)¡ Á: (17)

15In the moral hazard literature, uncertainty is usually assumed. E®ort is what an agentspends on improving the probability of a good state. By observing what state is going tohappen, it is di±cult to verify what e®ort an agent has chosen. However, if the preferencesof the agent are known publicly, other agents know what the amount of e®ort the agent willchoose. In the present paper, we avoid assuming uncertainty because we want to simplifyour analysis and because including uncertainty will not change the results qualitatively.See Bruce and Wong (1995) for a relevant model of moral hazard and insurance.16In the present model without uncertainty, the agreement between the two ¯rms may

include a requirement concerning the level of ¯xed cost, which is measurable and requiresa certain e®ort made by the local ¯rm. We do not consider this option in the presentpaper because it is not the focus here. Rather, we want to examine how moral hazardmay a®ect the formation of a joint venture.

10

Note that even though the optimal output of the joint venture is not a func-tion of the e®ort chosen by the home ¯rm, £̂(e) is.In the second stage, the two ¯rms share the nominal pro¯t in a Nash

bargaining game. In such a game, we assume that the alternative to a jointventure is a duopolistic situation as described above. In the latter situation,the local and foreign ¯rms will receive nominal pro¯ts equal to µd and µ¤d;respectively. As a result, the home ¯rm's share of the pro¯t of the IJV, s; ischosen to solve the following problem:

maxs

hs£̂(e)¡ µd

i h(1¡ s)£̂(e)¡ µ¤d

i: (18)

The ¯rst-order condition is equal to

s£̂(e) =£̂(e)¡ µ¤d + µd

2: (19)

Note that the left-hand side of (19) is the portion of the joint venture'snominal pro¯t the home ¯rm receives, based on the e®ort chosen by thehome ¯rm. Denote the share that satis¯es (19) by s(e); which is a functionof the home ¯rm's e®ort.In the ¯rst stage of the game, the home ¯rm chooses its e®ort to maximize

its net pro¯t. Its problem is

maxe

¼(e) = s(e)£̂(e)¡ g(e): (20)

Making use of (19), (20) reduces to

maxe

¼(e) =£̂(e)¡ µ¤d + µd

2¡ g(e): (21)

Using condition (17), the ¯rst-order condition is equal to

¡12f 0(e) = g0(e): (22)

The optimal e®ort can be illustrated in Figure 1. Schedule F0F0 is con-structed so that vertically it is half way between schedule FF and the horizon-tal axis. The equilibrium point is the intersecting point E0 between schedulesF0F0 and GG. A comparison between points E and E0 shows that with theformation of the joint venture, the home ¯rm chooses to spend less e®ort,

11

ev; which results in a higher \¯xed" cost of production f(ev), as comparedwith what it does when it produces alone or as a duopolist, ed. Denote theresulting pro¯t of the joint venture by £v ´ £̂(ev): Therefore the nominalpro¯ts of the home and foreign ¯rms are respectively equal to

s£v =£v ¡ µ¤d + µd

2(23)

(1¡ s)£v =£v + µ¤d ¡ µd

2: (24)

The net pro¯t of the home ¯rm is ¼v = s£v¡ g(ev): Because the foreign ¯rmdoes not have to spend any e®ort in forming the joint venture, its nominalpro¯t is the same as its net pro¯t, i.e., ¼¤v = (1¡ s)£v:

5 Choosing between Joint Venture and Sep-

arate Production

After explaining how a joint venture can be formed, the next question iswhether the ¯rms prefer to have a joint venture, or to produce the goodseparately as duopolists. Each ¯rm chooses the option that produces a higherpro¯t. Let us denote the outcome joint venture by V and the outcome ofseparate production by D. We then say that V Â D if both ¯rms prefer ajoint venture with each other to separate production, and that D Â V if atleast one ¯rm prefers separate production to a joint venture.We ¯rst analyze the home ¯rm.

5.1 The Home Firm

The home ¯rm prefers a joint venture over a duopoly production if and onlyif it gets a higher net pro¯t:

s£v ¡ gv > ¼d = µd ¡ gd: (25)

Making use of condition (23) and rearranging terms, (25) reduces to:

£v > µd + µ¤d ¡ 2(gd ¡ gv): (26)

Consider the following condition:

£v > µd + µ¤d: (A)

12

Condition (A) means that the net pro¯t of the joint venture is higher thanthe total net pro¯ts of the ¯rms when producing separately. From (22), weknow that gd > gv. Thus condition (A) implies (26).

5.2 The Foreign Firm

The foreign ¯rm prefers a joint venture if and only if its pro¯t share is largerthan the net pro¯t under duopoly competition:

(1¡ s)£v > ¼¤d = µ¤d ¡ g¤d: (27)

Making use of condition (24), (27) reduces to

£v > µd + µ¤d ¡ 2g¤d: (28)

It is easy to see that condition (A) implies (28).A joint venture will be formed if and only if both ¯rms prefer the joint

venture to separate production. Thus we have the following proposition:

Proposition 1 (a) V Â D if and only if conditions (26) and (28) are sat-is¯ed. (b) V Â D if condition (A) is satis¯ed.

Since condition (A) is a su±cient condition for the formation of a jointventure, let us analyze it more carefully. It states that the nominal pro¯t ofthe joint venture is greater than the sum of the nominal pro¯ts of the ¯rmswhen producing separately as Cournot duopolists. Using the de¯nitions ofthese nominal pro¯ts, we have

£v ¡ µd ¡ µ¤d = [p(Xv)Xv ¡ c¤Xv ¡ f(ev)¡ Á]¡hp(qd)xd ¡ cxd ¡ f(ed)

i¡hp(qd)x¤d ¡ c¤x¤d ¡ f ¤(e¤d)

i=

hp(Xv)Xv ¡ c¤Xv ¡ f(ed)

i¡hp(qd)xd ¡ c¤xd ¡ f(ed)

i¡hp(qd)x¤d ¡ c¤x¤d ¡ f(ed)

i+ (c¡ c¤)xd

+hf ¤(e¤d)¡ f(ed)

i+hf(ed)¡ f(ev)

i¡ Á: (29)

Let us analyze condition (29). The individual terms on its RHS have thefollowing signs:

13

(a)hp(Xv)Xv ¡ c¤Xv ¡ f(ed)

i¡hp(qd)xd ¡ c¤xd ¡ f(ed)

i¡hp(qd)x¤d ¡ c¤x¤d ¡ f(ed)

i> 0;

(b) (c¡ c¤)xd > 0;(c)

hf ¤(e¤d)¡ f(ed)

i> 0; and

(d)hf(ed)¡ f(ev)

i< 0.

The term in (a) represents the di®erence between the maximum pro¯twhen two identical ¯rms produce jointly and the sum of their pro¯ts whenthey produce non-cooperatively, which is positive because a joint venture canbetter exploit the monopoly power. The signs of the terms in (b) and (c) aredue to assumptions II and I, respectively. The term in (d) is negative due tothe moral hazard problem as explained before. As a result, condition (A) isnot necessarily satis¯ed.

Proposition 2 If the cooperation cost is su±ciently small, the home ¯rmalways prefers an IJV to separate production. If the cooperation cost is suf-¯ciently small and if the drop in the home ¯rm's e®ort will be su±cientlysmall, then condition (A) is satis¯ed and V Â D:

Proof. To see the ¯rst statement, assume that the home ¯rm chooses thesame e®ort em when cooperating with the foreign, as it does as a monopolist(or duopolist). Then the joint pro¯t of both ¯rms must be higher becauseboth ¯rms act as a single monopolist and because both ¯rms have comple-mentary technologies. So the home ¯rm must get a higher pro¯t. If thehome ¯rm chooses the e®ort it supplies optimally, its pro¯t will be higher,or not lower. To see the second statement, note that under the given condi-tions term (d) and Á in (29) are su±ciently small, and the RHS is positive,implying condition (A). By Proposition 1, V Â D:

Proposition 3 V Â D if (i) condition (28) is satis¯ed and (ii) g¤d < (gd ¡gv):

Proof. Conditions (i) and (ii) imply condition (26). So both ¯rms prefer anIJV.

14

6 Technology Spillover and Break-Up of A

Joint Venture

So far, we have been focusing on a one-period model. Although the jointventure makes use of the advantages of both ¯rms, the technology of each¯rm remains unchanged. We now turn to a two-period model to analyze whyin some cases a joint venture is formed but breaks up in the future.We assume that in each of the two periods, the ¯rms have the option of

forming a joint venture under the conditions described above (V ), and thatof producing separately as duopolists (D): To represent di®erent outcomesin these two periods, we use a duplet (i; j); where i; j = V; D: In general,there are altogether four possible outcomes: (V; V ); (V; D); (D; V ); and (D;D):The main feature of the present two-period model is that when the ¯rms

form a joint venture, both are able to learn from each other. This leads to achange in the technology they use in the second period, thus a®ecting theirchoice between a joint venture and separate production.We now analyze more carefully the decision making of the ¯rms. First,

let us be more explicit about how ¯rms learn from each other. Recalling thatthe home ¯rm is more e±cient in terms of the \¯xed" cost of productionbut less e±cient in terms of the marginal cost, we make the following simpleassumptions:

(i) In the ¯rst period, both ¯rms have the technology described in previoussections.

(ii) In the second period, the home ¯rm has a marginal cost equal to ±c¤;and the foreign ¯rm faces a \¯xed" cost function of ±¤f(e) and has topay an e®ort cost of ±¤g(e) if it spends an e®ort of e¤; where 1 < ± < c=c¤

and 1 < ±¤ < [f ¤(e) + g¤(e)]=[f(e) + g(e)] for all relevent values of e:

(iii) Learning is free.

Assumption (ii) means that the home ¯rm learns from the foreign ¯rmthrough a joint venture, and is able to reduce its marginal cost althoughthe marginal cost is still greater than that of the foreign ¯rm. Similarly,the foreign ¯rm reduces its \¯xed" cost of production although the latter isgreater than what the home ¯rm has. The foreign ¯rm learns not only how

15

to change its \¯xed" cost but also the cost of its e®ort. Assumption (iii) ismade to simplify our analysis.Before we analyze the two-period model, let us examine how the tech-

nology spillovers may a®ect the two ¯rms. Suppose that after one period ofcooperation in the form of a joint venture, both ¯rms now want to competeas Cournot duopolists. This case can easily be compared with the one de-scribed in Section 3, except that the ¯rms' technologies are di®erent thanbefore. Our main focus is how the pro¯ts may change after the ¯rms havelearned from each other.We can again conceptually assume that the ¯rms compete in two separate

stages: In stage one, they choose the level of e®ort, and then in stage twothey compete in a Cournot way. Our analysis is simpli¯ed by the fact thatthe e®ort they choose is independent of the output. We thus can immediatelyget the following results: (a) Since the home ¯rm maintains the same \¯xed"cost of production and the same e®ort cost function as before, the home ¯rmchooses the same e®ort as before, i.e., ed: (b) Since the foreign ¯rm learnsfrom the home ¯rm in terms of the latter's \¯xed" cost and e®ort cost, the¯rst-order condition of an optimal e®ort as expressed in (6) also applies tothe foreign ¯rm. This means that the foreign ¯rm will also choose the samee®ort as the home ¯rm, ed:As a result, the marginal cost and ¯xed cost of the home ¯rm in the

present case are ±c¤ (< c) and f(ed) respectively, and those of the foreign

¯rm are c¤ and ±¤f(ed)³< f¤(ed)

´: The Nash equilibrium can be derived

using the technique described in Section 3 and can be compared with theCournot equilibrium before the technology spillover. Because the marginalcost of the foreign ¯rm is unchanged, its reaction function will remain thesame as before, which is illustrated by FF in Figure 2. For the home ¯rm,the reduction in marginal cost will encourage it to produce more for eachproduction level of the foreign ¯rm, meaning that its new reaction curve,such as H0H0 in Figure 2, is on the right of the original reaction curve, HH.Schedule H0H0 cuts schedule FF at the new Nash equilibrium point N0: Usinga tilde and a superscript \d" to represent variables after technology spillover,let us denote the new equilibrium output of the home ¯rm by ~xd and thatof the foreign ¯rm by ~x¤d: Comparing the new and initial Nash equilibriumpoints, we get ~xd > xd and ~x¤d < x¤d: This con¯rms conditions (12) and (13).How are the pro¯ts of the two ¯rms a®ected by the technology spillovers?

Let us consider ¯rst the home ¯rm. Because the e®ort it chooses is inde-

16

pendent of output level and thus is not a®ected by the spillover e®ect, thechange in its net pro¯t is equal to:

~¼d ¡ ¼d = [~pd~xd ¡ ±c¤~xd ¡ fd ¡ gd]¡ [pdxd ¡ cxd ¡ fd ¡ gd]= [(~pd~xd ¡ c~xd ¡ fd)¡ (pdxd ¡ cxd ¡ fd)] + (c¡ ±c¤)~xd> 0; (30)

where the sign is due to the following factors:

(a) [(~pd~xd ¡ c~xd ¡ fd) ¡ (pdxd ¡ cxd ¡ fd)] > 0; because of a drop in theoutput of its rival; and

(b) (c¡ ±c¤)~xd > 0; because of the spillover e®ect.

We now turn to the foreign ¯rm. The change in its pro¯t is equal to

~¼¤d ¡ ¼¤d = [~pd~x¤d ¡ c¤~x¤d ¡ ~f ¤d ¡ ~g¤d]¡ [pdx¤d ¡ c¤x¤d ¡ f ¤d ¡ g¤d]= (~pd~x¤d ¡ c¤~x¤d)¡ (pdx¤d ¡ c¤x¤d)

+ (¡ ~f ¤d ¡ ~g¤d)¡ (¡f ¤d ¡ g¤d); (31)

where ~f ¤d = ±¤f(ed) and ~g¤d = ±¤g(ed). In analyzing (31), note that we have

(¡ ~f¤d ¡ ~g¤d)¡ (¡f ¤d ¡ g¤d)= [¡±¤f(ed)¡ ±¤g(ed)]¡ [¡f ¤(e¤d)¡ g¤(e¤d)]= f[¡±¤f(ed)¡ ±¤g(ed)]¡ [¡±¤f(e¤d)¡ ±¤g(e¤d)]g

+ f[¡±¤f(e¤d)¡ ±¤g(e¤d)]¡ [¡f¤(e¤d)¡ g¤(e¤d)]g:

To determine the sign of the change in pro¯t given in (31), let us note thefollowing factors:

(a) (~pd~x¤d¡c¤~x¤d)¡(pdx¤d¡c¤x¤d) < 0; because of an increase in the home¯rm's output;

(b) f[¡±¤f(ed)¡ ±¤g(ed)]¡ [¡±¤f(e¤d)¡ ±¤g(e¤d)]g > 0; due to pro¯t max-imization after the spillover; and

17

(c) f[¡±¤f(e¤d)¡ ±¤g(e¤d)]¡ [¡f ¤(e¤d)¡ g¤(e¤d)]g > 0; due to technologyspillover.

Combining these factors together, we note that the sign of the change inpro¯t given by (31) is ambiguous.We now turn to the case in which the two ¯rms choose to produce sepa-

rately as duopolists in period 1. In this case, there is no technology transfer,meaning that technologies of the ¯rms remain the same in period 2. Thismeans that (D, V ) will not be an outcome in the present model. The reasonis that if D Â V in period 1, then the ¯rm(s) that prefers separate productionwill still prefer separate production in period 2. Thus we are left with threepossible outcomes.In the present two-period model, denote the sum of the discounted pro¯ts

of the home ¯rm by Wij and that of the foreign ¯rms by W¤ij; where i = d; v

for the duopolist or joint venture outcome in period 1, respectively, and j = d;v for those in period 2, respectively. Also denote the discount rates of thehome and foreign ¯rms by ½ and ½¤; respectively. Therefore the two-periodpro¯ts of the ¯rms in the following three cases are:

1. outcome (D; D):

Wdd = (1 + ½)¼d (32)

W ¤dd = (1 + ½¤)¼¤d: (33)

2. outcome (V; V ):

Wvv = (1 + ½)¼v (34)

W ¤vv = (1 + ½¤)¼¤v: (35)

3. outcome (V; D):

Wvd = ¼v + ½~¼d (36)

W ¤vd = ¼¤v + ½¤~¼¤d: (37)

It should be noted that if the ¯rms form a joint venture in period 1, andif in period 2 the joint venture stays, the technology spillover will have noimpacts on the ¯rms' pro¯ts because the joint venture will still be using themarginal cost supplied by the foreign ¯rm and the \¯xed" cost supplied bythe home ¯rm. Let us now compare these three outcomes. Our comparisoncan be divided into three parts:

18

6.1 (V; V ) versus (D; D)

Conditions (32) and (33) indicate that the two-period pro¯t of each ¯rm isdirectly proportional to its pro¯t in one period. Therefore (V; V ) Â (D; D)if and only if V Â D: See Propositions 1 and 2.The options of the home ¯rm can be illustrated in Figure 3, where the

horizontal axis represents its (current value) net pro¯t in period 1, ¼1; and thevertical axis stands for its net pro¯t in period 2, ¼2: Di®erent combinationsof ¼1 and ¼2 that yield the same two-period pro¯t of the home ¯rm, Wdd =¼1 + ½¼2; for various values of Wdd can be represented by parallel iso-pro¯tlines with a slope of ¡1=½: Line AB is an iso-pro¯t line passing through thepoint (¼d; ¼d): This line stands for the two-period pro¯t with outcome (D;D): The net pro¯ts of the home ¯rm under outcome (V; V ) can be representedby a point on the 45±-line OHK. Therefore the home ¯rm prefers (V; V ) to(D; D) if and only if its net pro¯ts under outcome (V; V ) is on the linesegment HK above H.The choice of the foreign ¯rm can be illustrated in the same way. Figure

4 represents the ¯rm's period-1 net pro¯t ¼¤1 (horizontal axis) and period-2net pro¯t ¼¤2 (vertical axis). CD is the iso-pro¯t line with a slope of ¡1=½¤and passing through point F (¼¤d; ¼¤d): OFG is the 45±-line from the origin.So the foreign ¯rm prefers (V; V ) to (D; D) if and only if its net pro¯ts underoutcome (V; V ) is on the line segment FG above F.

6.2 (V; D) versus (D; D)

The question now is, if the ¯rms are going to produce separately as duopolistsin the second period, does it make any sense to form a joint venture in the¯rst period? To answer this question, let us recall that in the presence ofa joint venture, both ¯rms learn from each other and are able to reducetheir costs. However, the two ¯rms are a®ected di®erently by the technologyspillovers. As explained earlier, the joint venture allows the home ¯rm toachieve a higher pro¯t in the second period, but the foreign ¯rm may be hurtby the joint venture. The impacts of the joint venture have to be taken intoaccount when comparing between these two outcomes.The pro¯ts of the home ¯rm when a joint venture is formed in the ¯rst

period can be represented by a point in Figure 3, (¼v; ~¼d): As explainedearlier, ~¼d > ¼d: Three possible points representing possible pro¯ts of the

19

home ¯rm are shown in the diagram, L, M, and N, all of which depict ahigher second-period pro¯t than ¼d. Point L represents a lower two-periodpro¯t for the home ¯rm than under (D; D): Thus (V; D) is dominated by(D; D). At point N, the home ¯rm gets a higher pro¯t in each of the periodunder the outcome (V; D); and thus prefers it to the outcome (D; D): PointM shows an interesting case. This point means that the formation of a jointventure lowers the current value of the home ¯rm's pro¯t, ¼v < ¼d: However,because point M is above line AB, the home ¯rm is able to gain enough inthe second period through learning from the foreign ¯rm to compensate forthe drop in its pro¯t in the ¯rst period. We call this type of cooperationwith the foreign ¯rm a strategic cooperation. The shaded region labeled SC(and also called region b) in Figure 3 represents possible pro¯ts with whicha strategic cooperation with the foreign ¯rm is preferred by the home ¯rm.We now turn to the foreign ¯rm. Figure 4 represents the pro¯t of the ¯rm

in each period under the outcome (D; D); ¼¤d; and the iso-pro¯t line CD.The pro¯ts of the ¯rm under the outcome (V; D) can also be represented inthe ¯gure. Following the analysis above, we can argue that the ¯rm prefers(V; D) to (D; D) if and only if the point for the pro¯ts under (V; D) is aboveline CD.Two regions in Figure 4 deserve more analysis. Region labeled SC (also

labeled region b), which is shaded, represents the pro¯ts of the foreign ¯rmwith which it prefers (V; D) to (D; D) even though it gets a pro¯t from ajoint venture in period 1 less than what it gets if it produces separately. Thisis what we call strategic cooperation. Another region of interest is labeledSN (also labeled region f) and is shaded. In this region, the foreign ¯rmreceives a pro¯t in period 1 from a joint venture higher than that it pro-duces separately as a duopolist. However, forming a joint venture causes abig drop in the pro¯t in period 2 when both ¯rms compete in separate pro-duction as compared with what it can get in the same period should therebe no cooperation. The cooperation in period 1 is thus too costly to theforeign ¯rm, which then prefers not to cooperate. We call this case strategicnoncooperation.

6.3 (V; V ) versus (V; D)

We now turn to another case: both ¯rms cooperate in period 1 but in period2 they have the option of continuing the cooperation or breaking up the jointventure. For this case, the choice boils down to what the ¯rms can get in

20

period 2 in these two options: joint venture and duopolistic production. Suchcomparison has been done earlier, except that the two frms have improvedtheir technologies. For example, in period 2 the home ¯rm chooses to keepthe joint venture if and only if the following condition holds:

¼v > ~¼d; (38)

or, applying condition (26), if and only if

£v > ~µd+ ~µ

¤d ¡ 2(~gd ¡ gv): (39)

The decision criterion for the home ¯rm can also be illustrated in Figure3. Note that its net pro¯ts in the two periods under outcome (V; V ) arerepresented by a point on line OHK. Therefore condition (38) means thatthe net pro¯ts (¼v; ~¼d) under outcome (V; D) are depicted by a point belowline OHK.Similarly, in period 2 the foreign ¯rm prefers to keep the joint venture if

and only if¼¤v > ~¼¤d; (40)

or, applying condition (28), if and only if

£v > ~µd+ ~µ

¤d ¡ 2~g¤d: (41)

Using the same argument as for the home ¯rm, we can conclude that theforeign ¯rm chooses to keep the joint venture if its net pro¯ts (¼¤v; ~¼¤d)under outcome (V; D) is represented by a point below line OFG in Figure 4.

6.4 The Optimal Choice

So far we have been comparing only any two of the outcomes. We nowconsider these three options together. We ¯rst consider the home ¯rm, andsummarize its decisions by the following rules:

(i) The home ¯rm prefers (V; V ) to (D; D) if and only if (¼v; ¼v) is online segment HK and above H.

(ii) The home ¯rm prefers (V; D) to (D; D) if and only if (¼v; ~¼d) is aboveAB.

21

(iii) The home ¯rm prefers (V; V ) to (V; D) if and only if (¼v; ~¼d) is belowOHK.

In Figure 3, we identify four regions a; b; c; and d; and try to determinehow the decision of the home ¯rm will be a®ected if (¼v; ~¼d) is representedby a point in one of these regions. Note that we neglect the area below ahorizontal line passing through point H, because if a joint venture is formedin period 1, the pro¯t of the home ¯rm in period 2 will be higher than¼d through technology spillover. Based on the three points in the abovesummary, we can make the following conclusion concerning the choice of the¯rm:

(a) The home ¯rm will choose (V; D) if (¼v; ~¼d) is in region b or c: To seewhy, note that in one of these regions, (¼v; ~¼d) is above AB and OHK.So by rules (ii) and (iii), (V; D) dominates. Note that region b is theregion of strategic cooperation.

(b) The home ¯rm will choose (V; V ) if (¼v; ~¼d) is in region d: In this case,(¼v; ¼v) is on line segment HK and above H and (¼v; ~¼d) is below OHK.The result follows from rules (i) and (iii).

(c) The home ¯rm will choose (D; D) if (¼v; ~¼d) is in region a: In thiscase, (¼v; ¼v) is not on line segment HK and (¼v; ~¼d) is below AB. Theresult follows from rules (i) and (ii).

The above results mean that if (¼v; ~¼d) is in region b; c; or d; the home¯rm prefers to have a joint venture in period 1. If it is in region b or c; thenthe ¯rm prefers to break away from the joint venture in period 2.We can now turn to the foreign ¯rm. Rules similiar to rules (i) to (iii)

given above can be stated for the ¯rm in its decision. The decision criterioncan also be stated in terms of the location of the net pro¯t point (¼¤v; ~¼¤d):One di®erence between these two ¯rms is that with duopolistic productionin period 2 the foreign ¯rm may experience a drop in its net pro¯t if a jointventure is formed in period 1. As a result, the region below a horizontal linethrough point F cannot be excluded for a point representing (¼¤v; ~¼¤d): InFigure 4, regions a to h can be identi¯ed.

22

(a) The foreign ¯rm will choose (V; D) if (¼¤v; ~¼¤d) is in region b or c: Inone of these regions, (¼¤v; ~¼¤d) is above CD and OFG. Note that regionb is the region of strategic cooperation.

(b) The foreign ¯rm will choose (V; V ) if (¼¤v; ~¼¤d) is in region d or e: (¼¤v;¼¤v) is on line segment FG and above F and (¼¤v; ~¼¤d) is below OFG.

(c) The foreign ¯rm will choose (D; D) if (¼¤v; ~¼¤d) is in regions a; f; g orh: In this case, (¼¤v; ¼¤v) is not on line segment FG and (¼¤v; ~¼¤d) isbelow CD. Note that region f is the region of strategic non-cooperation.This means that in this region, the foreign ¯rm will not choose to havea joint venture and then a break up in period 2, but it will prefer tomaintain separate production throughout the two periods.

Let us now bring both ¯rms together and ¯nd out the actual outcome indi®erent cases. The following rules can be used to determine the outcome:

1. If both ¯rms prefer the same outcome, that outcome will happen.

2. If any one of the ¯rms prefers (D; D); a joint venture will not be formed,and (D; D) must be the outcome.

3. If a ¯rm prefers (V; V ) while the other ¯rm wants (V; D); then (V;V ) will not be formed and the former ¯rm will choose between (V; D)and (D; D): Such decision of the former ¯rm will dominate and it willbecome the actual outcome.

Using the above the rules, the outcomes are given in Table 1.

Table 1: Outcomes in Di®erent Cases:

H (V; V ) (V; D) (D; D)F(V; V ) (V; V ) ? (D; D)

(V; D) (V; D) (V; D) (D; D)

(D; D) (D; D) (D; D) (D; D)

23

Note: ? means that the outcome is (V; D) if (¼¤v; ~¼¤d) is in region e in Figure4, or is (D; D) if (¼¤v; ~¼¤d) is in region f .

In Table 1, the columns show di®erent options the home ¯rm preferswhile the rows give those of the foreign ¯rm. The contents of the table canbe constructed based on rules (1) to (3). Note that if the foreign ¯rm prefers(V; V ) while the home ¯rm wants (V; D); the foreign ¯rm is left with theoptions of (V; D) and (D; D): What it chooses depends on the location of(¼¤v; ~¼¤d) in Figure 4. If, on the other hand, the home ¯rm prefers (V; V )while the foreign ¯rm wants (V; D); the home ¯rm will choose (V; D) fromthe remaining options because if a joint venture is formed in period 1, it canhave a higher pro¯t when producing as a duopolist in period 2 than what itcan have in the same period if no joint venture is formed.

7 Concluding Remarks

In this paper, we provided a theory to explain why a MNC and a local ¯rmchoose to form an IJV and why they decide to dissolve it sometime later. Ourtheory has two important features: moral hazard and technology spillover.The presence of moral hazard means that an IJV may not be bene¯cial toboth ¯rms, and so there are cases in which at least one of the ¯rms prefersnot to have a joint venture. Technology spillover means that in the presenceof a joint venture the ¯rms learn from each other and are able to improvetheir technology. It is the technology spillover that is the key in our theoryto the formation of an IJV and its breakup in the future.In this paper, we introduced the concepts of strategic cooperation and

strategic noncooperation. Strategic cooperation exists if one of the ¯rms iswilling to sacri¯ce in the short run in forming a joint venture if the jointventure can bring a bigger future gain. Strategic noncooperation, on theother hand, is the case when a ¯rm chooses not to form a joint venture eventhough it will bring a short-term bene¯t. This paper shows that the usualone-period analysis of joint ventures may give misleading results.

24

O e

Figure 1

G

G

F

F’

E’

E

F’

F

edev

-f and g’’

O x

Figure 2

’

’

x*

N

N

H

H

H

F

F

xdxd ~

x d~*

x d*

’H

O x

Figure 2

’

’

x*

N

N

H

H

H

F

F

xdxd ~

x d~*

x d*

’H

’H’H’H’H

O

Figure 3

�1�d

�d

. . ..

L N

SC

�2

A

B

Ha

c

d

b

M

K

O

Figure 4

�1

.

SC

�2

�d*

�d*

SN

F

D

C

G

a

c

d

e

g

h

b

f

References

[1] Beamish, Paul W. (1985), \The Characteristics of Joint Ventures in De-veloped and Developing Countries," Columbia Journal of World Busi-ness, Fall, 13{19.

[2] Benito, Gabriel R.G. and Geir Gripsrud (1992), \The Expansion ofForeign Direct Investments: Discrete Rational Location Choices or aCultural Learning Process?", Journal of International Business Stud-ies, Third Quarter, 461{476.

[3] Blodgett, Linda Longfellow (1991), \Partner Contributions as Predic-tors of Equity Share in International Joint Ventures," Journal of Inter-national Business Studies, First Quarter, 63{79.

[4] Bruce, Neil, and Kar-yiu Wong (1996). \Moral Hazard, MonitoringCosts, and Optimal Government Intervention," Journal of Risk and Un-certainty, 12: 77{90.

[5] Chowdhury, Jafor (1992), \Performance of International Joint Venturesand Wholly Owned Foreign Subsidiaries: A Comparative Perspective,"Management International Review, 32: 115{133.

[6] Franko, Lawrence G. (1971), Joint Venture Survival in MultinationalCorporations, Praeger.

[7] Geringer, J. Michael and Louis Hebert (1991), \Measuring Performanceof International Joint Ventures," Journal of International BusinessStudies, 22: 249{263.

[8] Gomes-Casseres, Benjamin (1987), \Joint Venture Instability: Is It aProblem?", Columbia Journal of World Business, Summer, 97{102.

[9] Gomes-Casseres, Benjamin (1989), \Joint Ventures in the Face of GlobalCompetition," Sloan Management Review, Spring, 17{26.

[10] Hennart, Jean-Francois (1988), \A Transactin Costs Theory of EquityJoint Ventures," Strategic Management Journal, 9: 361{374.

25

[11] Kogut, Bruce (1988a), \Joint Ventures: Theoretical and Empirical Per-spectives," Strategic Management Journal, 9: 319{332.

[12] Kogut, Bruce (1988b), \A Study of the Life Cycle of Joint Ventures,"Management International Review, Special Issue, 39{52.

[13] Kogut, Bruce and Harbir Singh (1988), \The E®ect of National Cul-ture on the Choice of Entry Mode," Journal of International BusinessStudies, 19: 411{432.

[14] Lee, Chol and Beamish, Paul W. (1995), \The Characteristics and Per-formance of Korean Joint Ventures in LDCs," Journal of InternationalBusiness Studies, Third Quarter, 637{654.

[15] Leung, Wing-Fai (1998), \A Model of Coexistence of InternationalJoint Ventures and Foreign Wholly-Owned Subsidiaries," Japan and theWorld Economy, 10: 233{252.

[16] Leung, Wing-Fai (1997), \The Duration Times of International JointVentures and Foreign Wholly-Owned Subsidiaries," Applied Economics,29: 1255{1269.

[17] Miller, Robert, Jack Glen, Fred Jaspersen, and Yannis Karmokolias(1997). \International Joint Ventures in Developing Countries", Finance& Development, 26{29.

[18] Parkhe, Arvind (1991), \Inter¯rm Diversity, Organizational Learning,and Longevity in Global Strategic Alliance," Journal of InternationalBusiness Studies, Fourth Quarter, 579{601.

[19] Perlmutter, Howard V. and David A. Heenan (1986), \Cooperate toCompete Globally," Harvard Business Review, March-April, 136{152.

[20] Reynolds, John I. (1984), \The `Pinched Shoe' E®ect of InternationalJoint Ventures," Columbia Journal of World Business, Summer, 23{29.

[21] Wong, Kar-yiu (1995). International Trade in Goods and Factor Mobil-ity. Cambridge, Mass.: MIT Press.

26